Abstract Dryland ecosystems play an important role in the global carbon cycle, including regulating the inter-annual global carbon sink. Dynamic global vegetation models (DGVMs) are essential tools that can help us better understand carbon cycling in different ecosystems. Currently, there is limited knowledge of the performance of these models in drylands partly due to characterizing the heterogeneity of the vegetation and hydrometeorological conditions. The aim of this study is to evaluate the performance of a DGVM for drylands to facilitate improved understanding of gross primary production (GPP) as one of the important components of the carbon cycle. We performed a sensitivity analysis and calibrated the Ecosystem Demography (EDv2.2) DGVM to simulate GPP in a dryland watershed (Reynolds Creek Experimental Watershed, Idaho) in the western US for the years 2000-2017. GPP capacity and activity were investigated by comparing model simulations with GPP estimated from eddy covariance data (available from 2015-2017) and remote sensing products (2000-2017). Our results show good performance of EDv2.2 at daily timesteps ( R M S E ≈ 0.38 [ kgC / m 2 / year ] ) between simulated and measured GPP in lower elevations of the watershed. Moreover, remote sensing analysis show that EDv2.2 captures the long-term trends in this ecosystem and performs relatively well in capturing phenometrics (start/end of the season). The performance of the model degrades in more productive sites with greater GPP (located at higher elevations in the watershed). To improve model performance, future studies will need to introduce additional plant functional types for drylands such as our study area, and modify plant processes (e.g., plant hydraulics and phenology) in the model.

中文翻译：

---

生态系统人口学模型 (EDv2.2) 在模拟旱地研究区总初级生产能力和活动中的表现

摘要 旱地生态系统在全球碳循环中发挥着重要作用，包括调节全球年际碳汇。动态全球植被模型 (DGVM) 是必不可少的工具，可以帮助我们更好地了解不同生态系统中的碳循环。目前，对这些模型在旱地的性能了解有限，部分原因是表征了植被和水文气象条件的异质性。本研究的目的是评估旱地 DGVM 的性能，以促进更好地理解初级生产总值 (GPP) 作为碳循环的重要组成部分之一。我们进行了敏感性分析并校准了生态系统人口学 (EDv2.2) DGVM 以模拟旱地流域（雷诺兹溪实验流域，爱达荷州）在美国西部 2000-2017 年。通过将模型模拟与根据涡度协方差数据（2015-2017 年可用）和遥感产品（2000-2017 年）估计的 GPP 进行比较，研究了 GPP 容量和活动。我们的结果显示 EDv2.2 在流域较低海拔的模拟和测量 GPP 之间的每日时间步长（RMSE ≈ 0.38 [kgC / m 2 / 年]）的良好性能。此外，遥感分析表明，EDv2.2 捕捉了该生态系统的长期趋势，并且在捕捉表型（季节开始/结束）方面表现相对较好。该模型的性能在 GPP 较大的高产地点（位于流域中较高的海拔）中会降低。为了提高模型性能，